FU­TURE OF THE FU­TURE

En­ter­prise AI: work­ing smarter and harder on be­half of pro­fes­sion­als

Australian Transport News - - Operations + Strategy | Artificial Intelligence -

As it turns out, en­ter­prises across in­dus­tries share sim­i­lar needs where AI can aug­ment the pro­duc­tiv­ity and ef­fec­tive­ness of in­di­vid­ual knowl­edge work­ers.

As more com­pa­nies in more in­dus­tries em­brace AI, they will be­gin to use it to aug­ment core busi­ness ac­tiv­i­ties. Four key use cases shared across in­dus­tries are cus­tomer sup­port, ex­pert as­sist, in­put man­age­ment, and con­tent dis­cov­ery. • Cus­tomer sup­port: This refers to the au­to­ma­tion of cus­tomer in­ter­ac­tions by voice or chat­bots. In the en­ter­prise, these vir­tual as­sis­tants are be­ing de­vel­oped to al­low more com­plex di­a­logues with cus­tomers. In­dus­try re­search firm Gart­ner pre­dicts that, by 2020, the ma­jor­ity of com­mer­cial in­ter­ac­tions will take place be­tween cus­tomers and vir­tual agents. Though typ­i­cally met with scep­ti­cism from cus­tomers, the use of chat­bots and other vir­tual agents can be highly ef­fec­tive for au­tomat­ing low-level cus­tomer ser­vice in­quiries. Au­todesk, a global leader in com­puter-aided de­sign and en­gi­neer­ing soft­ware, has been im­mensely suc­cess­ful in this en­deav­our. Us­ing the IBM Wat­son con­ver­sa­tion ser­vice, the com­pany built the au­todesk vir­tual agent (AVA) that is able to an­swer 40 unique low-level queries at a rate of 30,000 in­ter­ac­tions per month and cuts re­sponse time to cus­tomer in­quiries from 1.5 days to five min­utes, or just about a 99 per cent re­duc­tion.

• Ex­pert as­sist: This refers to AI-based sys­tems that en­able knowl­edge work­ers to re­trieve and pro­duce in­for­ma­tion in a highly ef­fi­cient man­ner. Typ­i­cally, knowl­edge work­ers spend some­where be­tween 15 min­utes and more than an hour per day search­ing for busi­ness in­for­ma­tion in out­dated data­base sys­tems and cor­po­rate in­tranets pow­ered by key­word search tech­nol­ogy. These re­turn low-qual­ity re­sults in terms of the con­tent or re­spon­si­ble do­main ex­pert they are seek­ing. AI can re­duce this search time dra­mat­i­cally by us­ing au­to­matic clus­ter­ing, on­tolo­gies, and vis­ual-recognition tech­nolo­gies to iden­tify for the knowl­edge worker the right in­for­ma­tion, con­tent, and per­son. This re­duces search time by 50 per cent or more.

“AI can help the lo­gis­tics in­dus­try fun­da­men­tally shift its op­er­at­ing model”

• In­put man­age­ment: This refers to the au­to­matic (pre-) pro­cess­ing of in­com­ing mail, emails, in­voices, spread­sheets, pre­sen­ta­tions, PDFs, and other doc­u­ments with the help of AI. Most com­pa­nies need to process large vol­umes of in­for­ma­tion on a daily ba­sis. For ex­am­ple, the av­er­age knowl­edge worker sends and re­ceives more than 120 email mes­sages per day. Digi­ti­sa­tion helps to re­duce some of the bur­den: let­ters get scanned, in­voices are en­tered into ac­count­ing soft­ware, pre­sen­ta­tions and spread­sheets are up­loaded or ac­cessed in shared drives in the cloud. But be­fore this can hap­pen, hu­mans are typ­i­cally in­volved – some­one has to de­cide which depart­ment should re­ceive the let­ter, which text block should be used, or who else needs to be in­volved. AI-based so­lu­tions can do some of this (pre-) pro­cess­ing. VKB, an insurance com­pany in south­ern Ger­many, has im­ple­mented an AI-pow­ered in­put man­age­ment tool us­ing IBM Wat­son to iden­tify the top­ics and sen­ti­ment from un­struc­tured text in in­com­ing emails and let­ters. This in­for­ma­tion is used to pri­ori­tise and route these items to the cor­rect de­part­ments.

• Con­tent dis­cov­ery: This refers to the au­to­matic anal­y­sis of un­struc­tured data from emails, PDFs, pic­tures, au­dio and video, made pos­si­ble by the evo­lu­tion of big data an­a­lyt­ics tools and with the help of AI. Most com­pa­nies have lots of this un­struc­tured data – it typ­i­cally rep­re­sents 80 per cent of all com­pany data and usu­ally does not get an­a­lysed.

With the type of voice tech­nol­ogy of­fered by AVRL, sys­tems can in­ter­pret the se­man­tic mean­ing and in­tent of a phrase and then con­nect the vo­cal men­tion of prod­uct names with prod­uct in­for­ma­tion con­tained within an ERP, WMS, or TMS sys­tem. This al­lows lo­gis­tics op­er­a­tors to con­ver­sa­tion­ally in­ter­act with their IT sys­tems just as they would with an­other hu­man be­ing, even when us­ing col­lo­quial or in­for­mal phras­ing.

The abil­ity to au­to­mate in­put, store, and re­trieve in­for­ma­tion via con­ver­sa­tional voice in­ter­ac­tion re­moves time and com­plex­ity from many ware­house tasks that re­quire man­ual in­put or lookup of in­for­ma­tion.

PRE­DIC­TIVE LO­GIS­TICS

In a world char­ac­terised by un­cer­tainty and volatil­ity, AI can help the lo­gis­tics in­dus­try fun­da­men­tally shift its op­er­at­ing model from re­ac­tive ac­tions and fore­cast­ing to proac­tive op­er­a­tions with pre­dic­tive in­tel­li­gence.

This sec­tion will iden­tify both global, net­work-level pre­dic­tion op­por­tu­ni­ties as well as process-spe­cific pre­dic­tion op­por­tu­ni­ties. Pre­dic­tive net­work man­age­ment us­ing AI can sig­nif­i­cantly ad­vance the per­for­mance of lo­gis­tics op­er­a­tions.

For air freight, on-time and in- full ship­ment is crit­i­cal as it rep­re­sents only 1 per cent of global trade in terms of ton­nage but 35 per cent in terms of value.

Most air freight lanes and net­works are planned us­ing his­tor­i­cal data and ex­per­tise from pro­fes­sion­als with decades of in­dus­try ex­pe­ri­ence. DHL has de­vel­oped a ma­chine learn­ing-based tool to pre­dict air freight tran­sit time de­lays in or­der to en­able proac­tive mit­i­ga­tion.

By analysing 58 dif­fer­ent pa­ram­e­ters of in­ter­nal data, the ma­chine learn­ing model is able to pre­dict if the av­er­age daily tran­sit time for a given lane is ex­pected to rise or fall up to a week in ad­vance.

“The need for pre­dic­tive de­mand and ca­pac­ity plan­ning is self-ev­i­dent in the fid­get spin­ner boom of 2017”

Fur­ther­more, this so­lu­tion is able to iden­tify the top fac­tors in­flu­enc­ing ship­ment de­lays, in­clud­ing tem­po­ral fac­tors like de­par­ture day or op­er­a­tional fac­tors such as air­line on-time per­for­mance. This can help air freight for­warders plan ahead by re­mov­ing sub­jec­tive guess­work around when or with which air­line their ship­ments should fly.

The need for pre­dic­tive de­mand and ca­pac­ity plan­ning is self-ev­i­dent in the fid­get spin­ner boom of 2017. The three-pad­dle shaped spin­ning toy sud­denly and un­ex­pect­edly sold an es­ti­mated 50 mil­lion units in a pe­riod of sev­eral months.

In the US, fid­get spin­ners quickly shot up to 20 per cent of all re­tail toy sales in this pe­riod. This in­un­dated air freight and ex­press net­works with ship­ment vol­umes as toy mer­chants re­jected the nor­mal lead times as­so­ci­ated with ocean ship­ment of man­u­fac­tured goods.

The first videos of teenagers do­ing tricks with fid­get spin­ners be­gan trend­ing on YouTube in Fe­bru­ary 2017.

Hid­den deep within on­line brows­ing data, YouTube video views, and con­ver­sa­tions on so­cial me­dia, AI in its cur­rent state is able to iden­tify both the quan­ti­ta­tive rise in in­ter­est in a topic, as well as the con­text of that in­ter­est from se­man­tic un­der­stand­ing of un­struc­tured text. This en­ables pre­dic­tions to be made about which fads could boom in a sim­i­lar fash­ion to fid­get spin­ners.

Thanks to the speed and ef­fi­ciency of global sup­ply chains and ex­press net­works, even a few weeks’ lead time pro­vides sig­nif­i­cant ad­van­tage to mer­chants fac­ing un­ex­pected spikes in de­mand.

DHL’s Global Trade Barom­e­ter is a unique early in­di­ca­tion tool for the cur­rent state and fu­ture devel­op­ment of global trade. The tool uses large amounts of op­er­a­tional lo­gis­tics data, ad­vanced sta­tis­ti­cal mod­el­ling, and ar­ti­fi­cial in­tel­li­gence to give a monthly out­look on prospects for the global econ­omy.

The model takes a bot­tom-up ap­proach and uses im­port and ex­port data of in­ter­me­di­ate and early-cy­cle com­modi­ties from seven coun­tries to serve as the ba­sis in­put for the sys­tem, mea­sured in air freight and con­tainer­ised ocean freight lev­els.

Over­all, the sys­tem reg­u­larly eval­u­ates 240 mil­lion vari­ables from seven coun­tries (China, Ger­many, Great Bri­tain, In­dia, Ja­pan, South Korea, and the US) that rep­re­sent 75 per cent of global trade.

An AI en­gine, to­gether with other non-cog­ni­tive an­a­lyt­i­cal models, ex­presses a sin­gle value to rep­re­sent the weighted av­er­age of cur­rent trade growth and the up­com­ing two months of global trade.

Tests with his­tor­i­cal data re­veal a high cor­re­la­tion be­tween the DHL Global Trade Barom­e­ter and real con­tainer­ised trade, pro­vid­ing an ef­fec­tive three month out­look for global trade.

Pre­dic­tive risk man­age­ment is crit­i­cal for en­sur­ing sup­ply chain con­ti­nu­ity. The DHL

Re­silience360 plat­form is a cloud-based sup­ply chain risk-man­age­ment so­lu­tion that has been tai­lored to the needs of global lo­gis­tics op­er­a­tors.

For sup­ply chain lead­ers in many in­dus­tries, in­clud­ing the au­to­mo­tive, tech­nol­ogy, and en­gi­neer­ing and man­u­fac­tur­ing sec­tors, man­ag­ing the flow of com­po­nents from thou­sands of world­wide sup­pli­ers is a reg­u­lar part of daily busi­ness.

Prob­lems with sup­pli­ers, from ma­te­rial short­ages to poor labour prac­tices and even le­gal in­ves­ti­ga­tions, can cause crit­i­cal dis­rup­tions in the sup­ply chain. The Re­silience360 Sup­ply Watch mod­ule demon­strates the power of AI to mit­i­gate sup­plier risks.

Us­ing ad­vanced ma­chine learn­ing and nat­u­ral lan­guage pro­cess­ing tech­niques, Sup­ply Watch mon­i­tors the con­tent and con­text of 8 mil­lion posts from over 300,000 on­line and so­cial me­dia sources. This al­lows the sys­tem to un­der­stand the sen­ti­ment of on­line con­ver­sa­tions from un­struc­tured text to iden­tify in­di­ca­tors of risk ahead of time. This, in turn, al­lows sup­ply chain man­agers to take cor­rec­tive ac­tion ear­lier, and avoid dis­rup­tion.

In­tel­li­gent route op­ti­mi­sa­tion is crit­i­cal for lo­gis­tics op­er­a­tors to ef­fi­ciently trans­port, pick up, and de­liver ship­ments. Lo­gis­tics providers and last-mile de­liv­ery ex­perts typ­i­cally have deep ex­plicit and im­plicit knowl­edge of cities and their phys­i­cal char­ac­ter­is­tics. How­ever, new cus­tomer de­mands such as time-slot de­liv­er­ies, ad-hoc pick­ups and in­stant de­liv­ery are cre­at­ing new chal­lenges with in­tel­li­gent route op­ti­mi­sa­tion.

Deutsche Post DHL Group pi­o­neered the SmartTruck rout­ing ini­tia­tive in the early 2000s to de­velop pro­pri­etary real-time rout­ing al­go­rithms for its fleet op­er­a­tors and driv­ers.

Re­cently, new soft in­fra­struc­ture of cities, such as dig­i­tal and satel­lite maps, traf­fic pat­terns, and so­cial me­dia check-in lo­ca­tions are cre­at­ing a wealth of in­for­ma­tion that can aug­ment sys­tems like SmartTruck and im­prove the over­all rout­ing of truck driv­ers on de­liv­ery runs.

Satel­lite im­agery com­pany Dig­i­talGlobe de­liv­ers high-res­o­lu­tion pic­tures of the planet’s sur­face to ride-shar­ing gi­ant Uber. These im­ages pro­vide rich in­put sources for the devel­op­ment of ad­vanced map­ping tools to in­crease the pre­ci­sion of pick up, nav­i­ga­tion, and drop off be­tween its driv­ers and rid­ers.

Dig­i­talGlobe’s satel­lites can de­ci­pher new road-sur­face mark­ings, lane in­for­ma­tion, and street-scale changes to traf­fic pat­terns be­fore a city adds them to its of­fi­cial vec­tor map.

This level of de­tail from satel­lite im­agery can pro­vide valu­able new in­sight to plan­ning and nav­i­gat­ing routes not only for the trans­port of peo­ple but for ship­ments as well.

AI-POW­ERED CUS­TOMER EX­PE­RI­ENCE

The dy­namic be­tween lo­gis­tics providers and cus­tomers is chang­ing.

For most con­sumers, touch points with a lo­gis­tics com­pany be­gin at check­out with an on­line re­tailer and end with a suc­cess­ful de­liv­ery or some­times a prod­uct re­turn. For busi­nesses, touch points with lo­gis­tics providers are char­ac­terised by long-term ser­vice con­tracts, ser­vicelevel agree­ments, and the op­er­a­tion of com­plex global sup­ply chains. AI can help per­son­alise all of these cus­tomer touch points for lo­gis­tics providers, in­creas­ing cus­tomer loy­alty and re­ten­tion. Voice agents can sig­nif­i­cantly im­prove and per­son­alise the cus­tomer ex­pe­ri­ence with lo­gis­tics providers.

In 2017, DHL Par­cel was one of the first last-mile de­liv­ery com­pa­nies to of­fer a voice-based ser­vice to track parcels and pro­vide ship­ment in­for­ma­tion us­ing Ama­son’s Alexa. Cus­tomers with an Ama­zon Echo speaker in their home can sim­ply ask things like “Alexa, where is my par­cel?” or “Ask DHL, where is my par­cel.” Cus­tomers can then speak their al­phanu­meric track­ing num­ber and re­ceive ship­ment up­dates. If there is an is­sue with a ship­ment, cus­tomers can ask DHL for help and be routed to cus­tomer as­sis­tance.

Tak­ing this one step fur­ther, Is­raeli startup pack­age.ai has de­vel­oped Jenny, a con­ver­sa­tional agent to as­sist with

last-mile de­liv­ery. Jenny can con­tact par­cel re­cip­i­ents via Face­book Mes­sen­ger or SMS to co­or­di­nate de­liv­ery times, lo­ca­tions, and other spe­cialised in­struc­tions.

The chat-based ser­vice can also send driver progress up­dates and last-minute changes, as well as close the loop with de­liv­ery con­fir­ma­tion and gath­er­ing feed­back from cus­tomers.

The con­ver­sa­tional ca­pa­bil­i­ties and con­text Jenny is able to process makes for a nat­u­ral touch point for cus­tomers, as well as cut­ting down on up to 70 per cent of op­er­a­tional costs through route op­ti­mi­sa­tion and suc­cess­ful first-time de­liv­ery.

CON­TENT DIS­COV­ERY

AI can en­able lo­gis­tics com­pa­nies to be proac­tive about man­ag­ing their cus­tomer re­la­tion­ships. Al­ready to­day, hedge fund or­gan­i­sa­tions like Aidyia and Sen­tient Tech­nolo­gies are us­ing AI to ex­plore mar­ket data and make stock trades au­tonomously; each day, af­ter analysing ev­ery­thing from mar­ket prices and vol­umes to macroe­co­nomic data and cor­po­rate ac­count­ing doc­u­ments, their re­spec­tive AI en­gines make mar­ket pre­dic­tions and then “vote” on the best course of ac­tion.

Ini­tial tri­als showed a 2 per cent re­turn on an undis­closed sum which, while not sta­tis­ti­cally sig­nif­i­cant, rep­re­sents a sig­nif­i­cant shift in how firms can con­duct re­search and ex­e­cute trades.

An­tic­i­pa­tory Lo­gis­tics takes the AI-pow­ered lo­gis­tics cus­tomer ex­pe­ri­ence to the next level, de­liv­er­ing goods to cus­tomers be­fore they have even or­dered them or re­alised they needed them. An­tic­i­pa­tory lo­gis­tics seeks to lever­age the ca­pa­bil­i­ties of AI to an­a­lyse and draw pre­dic­tions from vast amounts of data such as brows­ing be­hav­iour, pur­chase his­tory, and de­mo­graphic norms as well as seem­ingly un­re­lated data sources such as weather data, so­cial me­dia chatter, and news re­ports to pre­dict what cus­tomers will pur­chase.

Ex­pos­ing these data sources to AI anal­y­sis, com­pa­nies can ef­fec­tively pre­dict de­mand and shorten de­liv­ery times by mov­ing in­ven­tory closer to cus­tomer lo­ca­tions and al­lo­cat­ing re­sources and ca­pac­ity to al­low for pre­vi­ously un­fore­seen de­mand. In some cases, it would even re­quire hav­ing non-pur­chased in­ven­tory con­stantly in tran­sit to al­low for in­stant de­liv­ery for an or­der placed while the goods are in mo­tion.

CON­CLU­SION

Ar­ti­fi­cial in­tel­li­gence is once again set to thrive; un­like past waves of hype and dis­il­lu­sion­ment, to­day’s cur­rent tech­nol­ogy, busi­ness, and so­ci­etal con­di­tions have never been more favourable to wide­spread use and adop­tion of AI.

In the con­sumer world, AI is al­ready here to stay. Among busi­nesses, lead­ing in­dus­tries such as tech, fi­nance, and, to a lesser ex­tent mo­bil­ity, are well into their AI jour­ney.

In­dus­trial en­ter­prise sec­tors like lo­gis­tics are be­gin­ning theirs in earnest now. Draw­ing on learn­ing from the con­sumer, en­ter­prise, re­tail, mo­bil­ity, and man­u­fac­tur­ing sec­tors pro­vides valu­able fore­sight of how AI

“En­ter­prise AI will al­le­vi­ate bur­den­some tasks that de­fine many as­pects of mod­ern work­ing life”

can be pro­duc­tively ap­plied in lo­gis­tics. En­ter­prise AI will al­le­vi­ate bur­den­some tasks that de­fine many as­pects of mod­ern work­ing life. As big data from op­er­a­tional, pub­lic, and pri­vate sources be­comes ex­posed to and pro­cessed by AI, the lo­gis­tics net­works will shift to a proac­tive and pre­dic­tive par­a­digm. Com­puter vi­sion and lan­guage-fo­cused AI will help lo­gis­tics op­er­a­tors see, un­der­stand, and in­ter­act with the world in novel, more ef­fi­cient ways than be­fore.

These same AI tech­nolo­gies will give rise to a new class of in­tel­li­gent lo­gis­tics as­sets that aug­ment hu­man ca­pa­bil­i­ties. In ad­di­tion, AI can help lo­gis­tics providers en­rich cus­tomer ex­pe­ri­ences through con­ver­sa­tional en­gage­ment, and de­liver items be­fore cus­tomers have even or­dered them.

AI, how­ever, it is not with­out its chal­lenges. The bias and in­tent of each AI de­vel­oper can be­come in­ter­twined in the sys­tem’s de­ci­sion-mak­ing func­tions, rais­ing com­plex ques­tions about the ethics of AI models. Here, busi­ness, so­ci­ety, and govern­ment bod­ies will need to de­velop stan­dards and reg­u­la­tions to en­sure the con­tin­ued progress of AI for the ben­e­fit of hu­man­ity.

We be­lieve the fu­ture of AI in lo­gis­tics is filled with po­ten­tial.

As sup­ply chain lead­ers con­tinue their dig­i­tal trans­for­ma­tion jour­ney, AI will be­come a big­ger and in­her­ent part of day-to-day busi­ness, ac­cel­er­at­ing the path to­wards a proac­tive, pre­dic­tive, au­to­mated, and per­son­alised fu­ture for lo­gis­tics.

Ul­ti­mately, AI will place a pre­mium on hu­man in­tu­ition, in­ter­ac­tion, and con­nec­tion al­low­ing peo­ple to con­trib­ute to more mean­ing­ful work.

In a 2017 sur­vey of CEOs, there was an al­most even four-way split among lead­ers who said they were us­ing AI, think­ing of us­ing AI, have heard of AI, or be­lieve AI is not a pri­or­ity.

This begs the ques­tion: who will teach our busi­ness lead­ers about AI?

DHL and IBM be­lieve the time for AI in lo­gis­tics is now. We look for­ward to hear­ing from you and cre­at­ing op­por­tu­ni­ties for col­lab­o­ra­tion and joint ex­plo­ration us­ing AI in your or­gan­i­sa­tion.

Be­low: Tax­on­omy of ma­chine learn­ing method­olo­gies

Above: AI in the in­ter­net of things

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.